Starting with Unitree GO2: GO2 lineup payload × price (domestic + overseas) × carrier/AMR integration reality and a starter plan (working draft)

⚠️ A working draft under continuous verification. Spec/price/integration links here are curl-tested reachable (HTTP 200). Four caveats first: (1) GO2 is consumer/education tier, haulage ~7–8 kg (regular/rated); the official "peak ~10–12 kg" is a peak, NOT continuous haulage — don't promise on it. (2) The official CN/EN spec tables list NO IP rating and NO "loaded vs unloaded" step/slope breakdown — so step/slope here are unloaded nominal. (3) RMB prices are media/e-commerce figures that fluctuate (the CN spec page shows no price); overseas per shop.unitree.com; specs per www.unitree.com (shop has price only). (4) Carrier integration: GO2 currently has NO ready-made AMR solution (see §3) — the most important conclusion here.

🌐 What this answers

The boss wants to start planning with the Unitree GO2. So this article focuses on GO2 only: lineup payload × price (domestic + overseas) × whether carrier/AMR integration is actually wired up, ending with a realistic "start with GO2" plan.

Following the siblings: Haulage selection tiers all models by payload; the Fusion article covers three fusion forms; Marsupial research covers marsupial research + cases. This one zooms in to the GO2 alone as a starter.

🐾 1. GO2 lineup payload & obstacle-crossing (by model)

The whole GO2 lineup shares the same body/weight/motors (~15 kg self-weight, ~45 N·m joint peak torque); the official site tiers only speed, slope angle, battery, compute; payload is given as a range (regular 7–8 kg / peak 10–12 kg), with no separate "continuous-walking vs standing-peak" field.

Reading notes:

• Haulage counts only the "regular 7–8 kg"; "peak 10–12 kg" is a ceiling — unreachable while walking, let alone climbing stairs with load.
• Step 15–16 cm, slope 30°/40°, GO2-W <70 cm obstacle are all unloaded nominal — the site gives no loaded figures; with load the real values drop (principle in Haulage selection · loaded-stair-climbing section).
• GO2-W wheeled is heavier (~18 kg) and crosses bigger obstacles (<70 cm), but its walking payload is the same tier (~8 kg); its edge is flat-ground rolling, not heavier loads.
• No IP rating on the official site — for oily/humid factory environments, confirm with Unitree in writing; don't assume ingress protection.

💰 2. Price (domestic official site + overseas channels)

• The CN spec page shows no price; exact RMB per Unitree's official JD/Tmall flagship in real time (the 9,997 / 18k here are media+e-commerce figures, re-check before purchase).
• Overseas dealers (Airpuria/RobotShop etc.) have finer pricing (Air ~$2,590, Pro ~$3,990, EDU ~$13k–16.5k) but are non-official, not listed as sources here.
• Key reminder: for secondary development/integration (next section), you must buy the EDU version — Air/Pro are cheaper but dev-locked; you can't plug them into an AMR.

🔗 3. Carrier / AMR integration reality (who has GO2 actually wired to?)

The most important honest conclusion: as of 2026-06, the Unitree GO2 has NO "ready-made" AMR/AGV system integration or car-dog deployment. Repeated CN/EN searches (vendors/universities/integrators/news/papers/videos/GitHub) found no public case of "GO2 mounted on an AGV/AMR" or "GO2 plugged into an AGV dispatch system." To build a carrier setup, you must self-build middleware on GO2's open SDK.

① Interfaces you can integrate with (open, but you write it)

② Only the EDU version truly allows secondary development (wrong version = can't plug in)

• CycloneDDS works out-of-box only on EDU; Air/Pro need custom firmware + root to use the SDK.
• Consumer Air/Pro are effectively dev-locked (home entertainment / intro coding); only EDU opens the full SDK incl. low-level torque/joint control (the prerequisite for custom gaits and host/AMR cooperation).
• Even on EDU, the community reports limits on low-level rt/lowcmd — confirm the open scope with Unitree against your real control needs before committing.

③ No VDA5050 support (and can't directly have it)

GO2 does not implement VDA5050 or other AGV dispatch protocols — and VDA5050 (MQTT+JSON) itself targets 2D wheeled AMRs only, not quadrupeds. So "scheduling GO2 inside Otsuka's existing AGV dispatch" can only be done via a self-written middleware bridge — no off-the-shelf protocol applies.

④ Ready accessories: mounts/arm exist, but no "dock-to-AGV" mechanism

• Unitree D1 6-axis arm (official, 500 g payload, EDU-only compatible)¹²
• GO2 modular mounting kit (third-party Picatinny rail for sensors/small load platforms)¹³
• ⚠️ No ready product for "GO2 mounting/dismounting an AGV / docking with an AMR" — this part must be self-developed (docking mechanism + dispatch middleware).

The frontier does have "quadruped × drone" marsupial carrier research, but that is not GO2 × wheeled AGV — don't conflate (see Marsupial research).

🎯 4. A realistic "start with GO2" plan for the boss

Combining the three sections, the realistic path is clear:

1. The version can only be GO2 EDU (on quote / overseas ~$13k tier) — Air/Pro, however cheap, are dev-locked, can't plug into an AMR, can't do a carrier setup. Don't buy Air/Pro to test integration.
2. Position: a low-cost PoC of ≤7 kg light load + stair-climbing + in-house dispatch software. GO2 can't haul heavy goods (regular 7–8 kg); its value is to validate the whole loop at lowest cost: small-item haulage + stairs + in-house path-dispatch/recognition software working end-to-end. Heavy goods go to B2/B2-W or carrier cooperation later.
3. Integration is necessarily "self-built" — and that's the differentiation. GO2 has no ready AMR docking; you write the "AMR↔GO2" dispatch middleware and docking logic yourself on unitree_sdk2/ROS2. That sounds like cost, but it's exactly the "software × hardware" moat in the business plan: buy the hardware, the cooperation orchestration is yours. Build this middleware on GO2 first; it carries over when you switch to B2/X30.
4. Phased:
   • Phase 1: one GO2 EDU validating the "stairs + light haulage + in-house software" loop (lowest failure cost).
   • Phase 2: self-build "(off-the-shelf) AMR + GO2 carrier docking + dispatch middleware," benchmarking UISEE × State Grid "car-dog integrated" for an in-plant prototype.
   • Phase 3: once cargo weight is confirmed, upgrade the heavy leg to B2/B2-W or a heavy AMR carrier, reusing the GO2-phase software stack.

Honest risks (state plainly to boss/client)

• GO2 can only do a light-load PoC, it can't carry heavy goods (≤7–8 kg) — use it to prove "the process works," not to promise haulage volume.
• Carrier/integration is all self-developed — GO2 has no ready AMR solution and no VDA5050; that's effort and opportunity, but don't sell it as "off-the-shelf."
• EDU version is mandatory + plan charging around real takt for factory IP/endurance (1–2 h, EDU long-life 2–4 h).

⚠️ To verify / to add

1. Confirm buying EDU: want me to compile a GO2 EDU Japan procurement channel + quote checklist (incl. EDU dev-permission items to confirm in writing)?
2. Real cargo weight/route: once I have it, I'll pin down the "start with GO2 → when you must switch to B2" upgrade threshold.
3. Want a minimal "AMR ↔ GO2 self-built middleware" architecture diagram (unitree_sdk2/ROS2 ↔ dispatch ↔ docking) as the Phase-1 PoC technical starting point?
4. Live RMB price: say the word and I'll check current Air/Pro/EDU listings on Unitree's official JD/Tmall store.